Aer Lingus needs to maintain flight schedules while minimizing the additional time required for deicing operations. Using historical weather data from the Irish meteorological service - Met Eireann, this project analyzes the relationship between rain and temperature to determine the likelihood of structural icing and the potential impact on deicing time. Based on the analysis, a Naive Model is recommended to create a deicing schedule that can minimize the additional time required for deicing and prevent flight delays.
Structural icing due to inclement weather can cause significant delays in the deicing process at Dublin Airport. To create a deicing schedule that minimizes the additional time required for deicing and prevents flight delays, this project analyzes historical weather data from Met Eireann to identify patterns in the relationship between rain, temperature, and holdover times. By analyzing these variables, the project aims to provide insights that can inform data-driven decisions and improve deicing schedules.
The dataset came from Kaggle. To download the dataset, please click here. All data was sourced from the Irish meteorological service - Met Eireann.
The project evaluates the performance of several machine learning models, including Naive Model, Random Walk, Random Walk with a drift, Autoregression, Moving Average, ARMA, and Neural Prophet. The Naive Model performs the best, particularly in predicting temperature.
Based on the analysis, it is recommended that Aer Lingus be prepared for shorter holdover times during the months of October to April. Additionally, it may be worth considering additional factors beyond just temperature in determining when to deice. While the Naive Model performed well, it is important to remain flexible and adaptable to changing weather conditions.
The project recommends that Aer Lingus use the Naive Model to create a deicing schedule that minimizes the additional time required for deicing and prevents flight delays. Additionally, the project suggests that other factors beyond just temperature, such as Type I deice, Type IV anti-ice, and flight scheduling, be considered in the deicing process.
To further test the accuracy of the model, it would be beneficial to evaluate the model's performance on a holdout set of data that it has not seen before. Additionally, it may be worth exploring other machine learning algorithms and techniques to potentially improve the accuracy of the model.
Presentation](https://github.com/casiasjc/Weather-Forecasting-for-Aviation-Operations/blob/main/Presentation.pdf) - Google slides of presentation Final_Clean_Notebook - The final notebook
To reproduce the results, please follow the instructions provided in the notebook.